Ambig-IaC: Multi-level Disambiguation for Interactive Cloud Infrastructure-as-Code Synthesis

2026-04-01

AI Summary

Natural language descriptions of cloud infrastructure requirements often suffer from ambiguity, limiting the accuracy of large language models in generating Infrastructure-as-Code (IaC) in a single pass. This work proposes a training-free, disagreement-driven framework that, for the first time, reveals IaC ambiguity to possess a composable three-layer structure—encompassing resources, topology, and attributes. By structurally decomposing requirements, the framework generates diverse candidate configurations, detects cross-layer disagreements, and actively poses high-information clarification questions to iteratively narrow the configuration space. The authors introduce Ambig-IaC, the first benchmark comprising 300 validated tasks, along with an evaluation protocol based on graph edit distance and embedding similarity. Experiments demonstrate that the proposed method achieves relative improvements of 18.4% and 25.4% in structural and attribute-level accuracy, respectively, significantly outperforming the strongest existing baselines.

Abstract

The scale and complexity of modern cloud infrastructure have made Infrastructure-as-Code (IaC) essential for managing deployments. While large Language models (LLMs) are increasingly being used to generate IaC configurations from natural language, user requests are often underspecified. Unlike traditional code generation, IaC configurations cannot be executed cheaply or iteratively repaired, forcing the LLMs into an almost one-shot regime. We observe that ambiguity in IaC exhibits a tractable compositional structure: configurations decompose into three hierarchical axes (resources, topology, attributes) where higher-level decisions constrain lower-level ones. We propose a training-free, disagreement-driven framework that generates diverse candidate specifications, identifies structural disagreements across these axes, ranks them by informativeness, and produces targeted clarification questions that progressively narrow the configuration space. We introduce \textsc{Ambig-IaC}, a benchmark of 300 validated IaC tasks with ambiguous prompts, and an evaluation framework based on graph edit distance and embedding similarity. Our method outperforms the strongest baseline, achieving relative improvements of +18.4\% and +25.4\% on structure and attribute evaluations, respectively.

Problem

• Infrastructure-as-Code
• ambiguity
• large language models
• cloud infrastructure
• natural language

Innovation

• Infrastructure-as-Code
• ambiguity disambiguation
• hierarchical decomposition
• clarification questioning
• LLM-based synthesis